aboutsummaryrefslogtreecommitdiff
path: root/config
diff options
context:
space:
mode:
authorAlex Auvolat <alex.auvolat@ens.fr>2015-07-02 11:15:37 -0400
committerAlex Auvolat <alex.auvolat@ens.fr>2015-07-02 11:18:45 -0400
commit3f3ab2bfe3ebfa266d433012be1c89c722d63352 (patch)
tree589915018911ec364dccb4b897ab108913be464f /config
parent32b078f28add3d22529e55aeac6674d924e9b510 (diff)
downloadtaxi-3f3ab2bfe3ebfa266d433012be1c89c722d63352.tar.gz
taxi-3f3ab2bfe3ebfa266d433012be1c89c722d63352.zip
Unify parameters for joint_simple_tgtcls_111_cswdtx_bigger{,_dropout}
Diffstat (limited to 'config')
-rw-r--r--config/joint_simple_mlp_tgtcls_111_cswdtx_bigger.py9
-rw-r--r--config/joint_simple_mlp_tgtcls_111_cswdtx_bigger_dropout.py7
-rw-r--r--config/memory_network_1.py1
3 files changed, 8 insertions, 9 deletions
diff --git a/config/joint_simple_mlp_tgtcls_111_cswdtx_bigger.py b/config/joint_simple_mlp_tgtcls_111_cswdtx_bigger.py
index 93ff5c7..8e991a1 100644
--- a/config/joint_simple_mlp_tgtcls_111_cswdtx_bigger.py
+++ b/config/joint_simple_mlp_tgtcls_111_cswdtx_bigger.py
@@ -29,14 +29,14 @@ dim_embeddings = [
# Common network part
dim_input = n_begin_end_pts * 2 * 2 + sum(x for (_, _, x) in dim_embeddings)
-dim_hidden = [1000]
+dim_hidden = [5000]
# Destination prediction part
-dim_hidden_dest = [400]
+dim_hidden_dest = [1000]
dim_output_dest = dest_tgtcls.shape[0]
# Time prediction part
-dim_hidden_time = [400]
+dim_hidden_time = [500]
dim_output_time = len(time_tgtcls)
# Cost ratio between distance cost and time cost
@@ -46,8 +46,7 @@ embed_weights_init = IsotropicGaussian(0.01)
mlp_weights_init = IsotropicGaussian(0.1)
mlp_biases_init = Constant(0.01)
-learning_rate = 0.000001
-momentum = 0.99
+# use adadelta, so no learning_rate or momentum
batch_size = 200
valid_set = 'cuts/test_times_0'
diff --git a/config/joint_simple_mlp_tgtcls_111_cswdtx_bigger_dropout.py b/config/joint_simple_mlp_tgtcls_111_cswdtx_bigger_dropout.py
index b047d7c..e1bd840 100644
--- a/config/joint_simple_mlp_tgtcls_111_cswdtx_bigger_dropout.py
+++ b/config/joint_simple_mlp_tgtcls_111_cswdtx_bigger_dropout.py
@@ -46,11 +46,10 @@ embed_weights_init = IsotropicGaussian(0.01)
mlp_weights_init = IsotropicGaussian(0.1)
mlp_biases_init = Constant(0.01)
-# apply_dropout = True
-# dropout_p = 0.5
+apply_dropout = True
+dropout_p = 0.5
-learning_rate = 0.001
-momentum = 0.9
+# use adadelta, so no learning_rate or momentum
batch_size = 200
valid_set = 'cuts/test_times_0'
diff --git a/config/memory_network_1.py b/config/memory_network_1.py
index 68e23bd..813c9d2 100644
--- a/config/memory_network_1.py
+++ b/config/memory_network_1.py
@@ -31,6 +31,7 @@ candidate_encoder.dim_hidden = [100, 100, 100]
candidate_encoder.weights_init = IsotropicGaussian(0.01)
candidate_encoder.biases_init = Constant(0.001)
+normalize_representation = True
embed_weights_init = IsotropicGaussian(0.001)